[MITgcm-devel] Fwd: [MITgcm-cvs] MITgcm/doc CVS Commit

Patrick Heimbach heimbach at MIT.EDU
Wed Jun 24 08:05:56 EDT 2009 

Martin,
heroic work!
I'll check the adjoint later today.
-p.

On Jun 24, 2009, at 5:12 AM, Martin Losch wrote:

> Hi Jean-Michel, Patrick and others,
>
> the transistion from the 3D to 2D fields UICE,VICE,HEFF, and AREA  
> is complete as far as a I can see. I have tried to fix the store  
> directives where I saw them (CADJ STORE area(:,:,:,bi,bj) -> CADJ  
> STORE area(:,:,bi,bj)) but I cannot guarantee that I didn't miss  
> one. As I pointed out to Patrick the main ad-problem should be with  
> advect.F and the additional fields uice/vice/heff/areaNm1 which do  
> not have store directives so far. The loop in advect.F has a  
> constant length 2.
>
> I have put the new fldNm1 (replaces gTmpFld) into a local common  
> block in seaice_advdiff.F but I don't think it's necessary. Maybe  
> Jean-Michel can remove if I am right.
>
> seaice+obcs: there so many different options, that I could not test  
> all of them, so it's very well possible that I have introduced a  
> problem that I did not notice afterwards, please be aware of that.
>
> I hope that the code is now easier to read/understand/modify with  
> the 3D-fields removed. And now I am going on vacation and leave you  
> with the mess alone (just kidding, but on the weekend I'll really  
> be gone for two weeks).
>
> Martin
>
> On Jun 23, 2009, at 8:16 PM, Jean-Michel Campin wrote:
>
>> Martin,
>>
>> I don't think gTmpFld was in a common block (or did i miss it ?).
>> But tmpFld needed to be in common block for S/R ADVECT exch.
>>
>> Jean-Michel
>>
>> On Tue, Jun 23, 2009 at 06:02:35PM +0200, Martin Losch wrote:
>>> I have now replace (so far only in my own copy) gTmpFld with  
>>> fldNm1 for
>>> the temporary array of the previous timestep for s/r advect. This  
>>> field
>>> is never exchanged, so it can probably be removed from the common  
>>> block?
>>>
>>> Martin
>>>
>>> On Jun 23, 2009, at 2:38 PM, Jean-Michel Campin wrote:
>>>
>>>> Hi Martin,
>>>>
>>>> I don't understand your concern about pickup.
>>>> I wrote in an email to the MITgcm-devel list on Wed, 17 Jun:
>>>>> Apart from that, I think we only store level1 in pickup-files.
>>>> so it should not be a problem, right ?
>>>>
>>>> There are other things that I don't follow very well (when I
>>>> looked to the code to fix the broken MPI+MTH, the copy in
>>>> tmpFld is also repeated inside S/R ADVECT; gtmpFlg is not
>>>> used to store a tendency, right ?).
>>>>
>>>> And regarding the adjoint, yes we are getting (in taf_ad.log) a  
>>>> lot of
>>>> TAF Warning, extensive recomputation (and variables you added
>>>> are also listed). Now on faulks, it takes about 1.h 15mn to run
>>>> each of the lab_sea test (compared to 30 to 35 second before).
>>>> And similar slow-down for offline_exf_seaice.
>>>>
>>>> We had in mind to make a tag at some point, but it's little bit
>>>> unfortunate that the seaice-pkg adjoint is not in its best shape.
>>>>
>>>> Cheers,
>>>> Jean-Michel
>>>>
>>>> On Tue, Jun 23, 2009 at 08:54:56AM +0200, Martin Losch wrote:
>>>>> Hi Patrich, not sure I can manage to do that today (too many  
>>>>> meetings
>>>>> and I am not feeling to well). Also I don't really know how to  
>>>>> do it
>>>>> right? Do I just replace the 3D fields by 2D fields in the
>>>>> seaice_read/
>>>>> write_pickup.F and that works?
>>>>> I'll give it a shot asap.
>>>>>
>>>>> Martin
>>>>> PS. did I break the adjoint? Most likely in advect.F
>>>>>
>>>>> On Jun 22, 2009, at 9:19 PM, Patrick Heimbach wrote:
>>>>>
>>>>>>
>>>>>> Hi Martin,
>>>>>>
>>>>>> J.M. and I were wondering whether you are trying
>>>>>> to avoid some important work (hence lots of work on pkg/seaice)
>>>>>> ;o)
>>>>>>
>>>>>> Since you're 1/2 way through with code modif's it might now be  
>>>>>> best
>>>>>> to complete the changes by removing the 3D fields altogether,
>>>>>> and I'll try to catch up with the adjoint after it's all done.
>>>>>> What do you think?
>>>>>>
>>>>>> Cheers
>>>>>> -p.
>>>>>>
>>>>>> Begin forwarded message:
>>>>>>
>>>>>>> From: Martin Losch <mlosch at mitgcm.org>
>>>>>>> Date: June 22, 2009 11:58:42 AM EDT
>>>>>>> To: mitgcm-cvs at mitgcm.org
>>>>>>> Subject: [MITgcm-cvs] MITgcm/doc CVS Commit
>>>>>>> Reply-To: MITgcm-cvs at mitgcm.org
>>>>>>>
>>>>>>> Update of /u/gcmpack/MITgcm/doc
>>>>>>> In directory forge:/tmp/cvs-serv27539/doc
>>>>>>>
>>>>>>> Modified Files:
>>>>>>> 	tag-index
>>>>>>> Log Message:
>>>>>>> document first steps in cleaning up the 3-time levels of
>>>>>>> UICE,VICE,HEFF,AREA:
>>>>>>> 1. introduce fields for time step (N-1 = *Nm1)
>>>>>>> 2. do not use levels 2 and 3 anymore (but they are still around)
>>>>>>>
>>>>>>> these changes probably break the adjoint, but the pickup is  
>>>>>>> still
>>>>>>> working, so I am stopping now to give P. a chance to fix the
>>>>>>> adjoint
>>>>>>> if necessary and then I will get rid of the 3D fields altogether
>>>>>>> carefully taking into account the pickup (not clear how).
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> MITgcm-cvs mailing list
>>>>>>> MITgcm-cvs at mitgcm.org
>>>>>>> http://mitgcm.org/mailman/listinfo/mitgcm-cvs
>>>>>>
>>>>>> ---
>>>>>> Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/ 
>>>>>> ~heimbach
>>>>>> MIT | EAPS 54-1518 | 77 Massachusetts Ave | Cambridge MA 02139  
>>>>>> USA
>>>>>> FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE  
>>>>>> patrick.heimbach
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> MITgcm-devel mailing list
>>>>>> MITgcm-devel at mitgcm.org
>>>>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>>>>
>>>>> _______________________________________________
>>>>> MITgcm-devel mailing list
>>>>> MITgcm-devel at mitgcm.org
>>>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>>> _______________________________________________
>>>> MITgcm-devel mailing list
>>>> MITgcm-devel at mitgcm.org
>>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>>
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

---
Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
MIT | EAPS 54-1518 | 77 Massachusetts Ave | Cambridge MA 02139 USA
FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach

More information about the MITgcm-devel mailing list